Interpretable machine learning model for new-onset atrial fibrillation prediction in critically ill patients: a multi-center study
Authors' conclusion
We developed a ML model to predict the risk of NOAF in critically ill patients without cardiac surgery and validated its potential as a clinically reliable tool.
Study context
- Population
- Critically ill ICU patients (MIMIC-IV for training; subset of MIMIC-III for external validation)
- Participants
- 16,528 MIMIC-IV patients
- Outcome
- New-onset AF after ICU admission
Methodological detailsmulticentre, machine learning, LASSO feature selection, SHAP interpretability
- Multicentre
- Machine learning
- LASSO feature selection
- SHAP interpretability
Performance
As reported in the abstract
Discrimination of the new-onset atrial fibrillation (NOAF) prediction model
| Test or model | AUC (validation) | AUC (external validation) |
|---|---|---|
| XGBoost | 0.891 (0.873-0.888) | 0.769 (0.756-0.782) |
XGBoost
AUC (validation)0.891 (0.873-0.888)
AUC (external validation)0.769 (0.756-0.782)
Further findings
Incidence of AF after ICU admission (MIMIC-IV)
- Developed AF post-ICU admission
- 1520 (9.2%)
Key predictors
- Risk probability defining high risk
- >0.6
- Age, mechanical ventilation, urine output, sepsis, blood urea nitrogen, SpO2, CRRT and weight
Parts of this Brief are withheld: the automatic checks could not confirm them against the abstract.
Original publication
Interpretable machine learning model for new-onset atrial fibrillation prediction in critically ill patients: a multi-center study. Crit Care. Published 29 October 2024. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →